Create the Edge Impulse Greengrass component

What you’ll create

You’ll create an AWS IoT Greengrass custom component that installs and runs the Edge Impulse Linux Runner service on your device. The component handles the Node.js and libvips prerequisites and manages the Edge Impulse Linux Runner lifecycle: install, run, and shutdown.

The component consists of two parts:

• Artifacts: Shell scripts, stored in Amazon S3, that install dependencies and launch the Edge Impulse Linux Runner.
• Recipe: A YAML file that tells Greengrass where to find the artifacts, what configuration to apply, and how to manage the component lifecycle.

Clone the component repository

Clone the Edge Impulse Greengrass components repository to get the recipe and artifact files:

    

        
        
git clone https://github.com/edgeimpulse/aws-greengrass-components.git

    

The repository contains the YAML recipe file and the shell script artifacts that you’ll upload to an S3 bucket.

Upload artifacts to Amazon S3

The Greengrass component downloads its artifacts from an S3 bucket at deployment time. You need to create a bucket and upload the shell scripts.

To upload artifacts to a bucket:

1. Open the AWS Console and navigate to S3.
2. Select Create bucket and give it a name (for example, my-ei-greengrass-artifacts):

Create an S3 bucket

3. Inside your new bucket, create the following directory structure:

       
   
           
           
   artifacts/EdgeImpulseServiceComponent/1.0.0/
   
       
   

4. Navigate to the 1.0.0 directory in your S3 bucket and select Upload. Upload all four files from the cloned repository’s ./artifacts/EdgeImpulseServiceComponent/1.0.0/ directory:

   • install.sh
   • run.sh
   • launch.sh
   • stop.sh

After the upload, your S3 bucket should look as follows:

Uploaded artifacts in S3

Customize the component recipe

The recipe YAML file tells Greengrass where to download the artifacts from S3. You need to update it with the name of your S3 bucket.

Open EdgeImpulseLinuxEdge Impulse Linux RunnerServiceComponent.yaml from the cloned repository and replace all occurrences of YOUR_S3_ARTIFACT_BUCKET with the name of your S3 bucket (for example, my-ei-greengrass-artifacts). After making updates, save the file.

Default configuration reference

The recipe file includes a default configuration JSON block. You don’t need to modify these defaults for the Learning Path — they’re overridden at deployment time by the device-specific JSON that you saved during hardware setup. However, understanding each field is useful for troubleshooting and customization.

    

        
        
{
   "node_version": "20.12.1",
   "vips_version": "8.12.1",
   "device_name": "MyEdgeImpulseDevice",
   "launch": "runner",
   "sleep_time_sec": 10,
   "lock_filename": "/tmp/ei_lockfile_runner",
   "gst_args": "__none__",
   "eiparams": "--greengrass",
   "iotcore_backoff": "5",
   "iotcore_qos": "1",
   "ei_bindir": "/usr/local/bin",
   "ei_sm_secret_id": "EI_API_KEY",
   "ei_sm_secret_name": "ei_api_key",
   "ei_ggc_user_groups": "video audio input users",
   "install_kvssink": "no",
   "publish_inference_base64_image": "no",
   "enable_cache_to_file": "no",
   "ei_poll_sleeptime_ms": 2500,
   "ei_local_model_file": "/home/ggc_user/data/currentModel.eim",
   "ei_shutdown_behavior": "__none__",
   "cache_file_directory": "__none__",
   "enable_threshold_limit": "no",
   "metrics_sleeptime_ms": 30000,
   "default_threshold": 50.0,
   "threshold_criteria": "ge",
   "enable_cache_to_s3": "no",
   "s3_bucket": "__none__"
}

    

The following table describes each configuration field:

Register the component in Greengrass

With the artifacts in S3 and the recipe updated, register the component in the AWS Console.

To register the component:

1. Navigate to AWS IoT Core > Greengrass > Components.
2. Select Create component.
3. Select Enter recipe as YAML as the input method.
4. Clear the default hello world YAML from the text box.
5. Copy and paste the entire contents of your edited EdgeImpulseLinuxEdge Impulse Linux RunnerServiceComponent.yaml file.
6. Select Create component.

Register the custom component

If the recipe format is valid and Greengrass can access the S3 artifacts, the component appears in your custom components list.

What you’ve accomplished and what’s next

You’ve cloned the Edge Impulse component repository and uploaded artifacts to S3. You’ve also customized the recipe with your bucket name, and registered the component in Greengrass.

Next, you’ll create a Greengrass deployment to push this component to your edge device.